基于YOLOv7和图像分块的车道线破损检测算法  

Lane line damage detection algorithm based on YOLOv7 and image block

作　　者：温王鹏 罗文婷 李林 张德津 陈文婷 吴镇涛 WEN Wangpeng;LUO Wenting;LI Lin;ZHANG Dejin;CHEN Wenting;WU Zhentao(College of Transportation and Civil Engineering,Fujian Agriculture and Forestry University,Fuzhou 350000,China;College of Transportation Engineering,Nanjing Tech University,Nanjing 211816,China;School of Architecture and Urban Planning,Shenzhen University,Shenzhen 518060,China)

机构地区：[1]福建农林大学交通与土木工程学院,福建福州350000 [2]南京工业大学交通运输工程学院,江苏南京211816 [3]深圳大学建筑与城市规划学院,广东深圳518060

出　　处：《传感器与微系统》2024年第9期131-134,139,共5页Transducer and Microsystem Technologies

基　　金：国家重点研发计划资助项目(2021YFB3202901);福建省高校产学合作重大项目(2020H6009)。

摘　　要：提出了一种结合YOLOv7和图像分块的车道线破损检测方法。首先,利用YOLOv7模型检测并提取车道线区域。其次,运用Otsu法计算每个子块的阈值及子块背景区域和目标区域的灰度均值差值,以此实现二值化。然后,采用双线性插值法平滑图像,实现车道线分割,并利用拓扑结构分析法提取车道线轮廓。最后,设计了像素统计、直线拟合、割断检测3种方法判断车道线是否破损。实验结果表明:在不同场景下,该算法在破损车道线检测中的精确率为91.79%,具有较好的检测效果和一定的应用价值。A lane line damage detection method combining YOLOv7 and image block is proposed.Firstly,YOLOv7 model is used to detect and extract the lane line area.Secondly,Otsu algorithm is used to calculate the threshold of each sub block and the gray mean difference between the background area and the target area in the sub block.The binarization is realized accordingly.Then,bilinear interpolation method is used to smooth the image and realize lane line segmentation,and the topological structure analysis method is used to extract the lane line contour.Finally,three methods including pixel statistics,straight line fitting and cutting detection are designed to judge whether the lane line is damaged.Experimental result shows that the precision of the algorithm is 91.79%in lane line damage detection under different scenarios,which has good detection effect and certain application value.

关 键 词：车道线破损检测 深度学习 YOLOv7算法 分块分割 最大类间方差法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] U491.5[自动化与计算机技术—计算机科学与技术]